使用DPG算法优化K-means初始聚类中心

Optimizing K-means Initial Cluster Centers Using DPG Algorithm

传统K-means算法具有速度快的优点,但是其初始聚类中心的随机性导致聚类结果不稳定、聚类精度较差。它的改进算法K-means++也有同样的问题。文章提出了一种基于密度峰值网格(Density Peak Grid,DPG)的K-means初始聚类中心优算法,该算法综合了网格分析的高效性和密度峰分析的准确性。在人工模拟数据集和UCI数据集上的实验结果表明,本文算法比传统的K-means和K-means++聚类效果更好,聚类速度更快。

The traditional K-means algorithm has the advantages of high speed,but the randomness of its initial cluster center leads to unstable clustering results and poor clustering accuracy.Its improved algorithm K-means++has the same problem.This paper proposes an optimal algorithm for K-means initial clustering center based on DPG,which combines the efficiency of grid analysis and the accuracy of density peak analysis.The experimental results on artificial simulation datasets and UCI datasets show that the proposed algorithm has better clustering effect and faster clustering speed than traditional K-means and K-means++.